Linköping University Medical Dissertations No. 1079
Relationships between Psychological Factors, Disability, Quality of Life and Health in Chronic Pain Disorders
Rehabilitation Medicine, Department of Clinical and Experimental Medicine, Faculty of Health Sciences Linköping University, SE-581 85 Sweden Linköping 2008
© Björn Börsbo Cover Design: Johan Werner Printed in Sweden by LiU-Tryck, Linköping, Sweden 2008. ISBN: 978-91-7393-797-9 ISSN 0345-0082
To my mother Karin
CONTENTS ABSTRACT ABBREVIATIONS ORIGINAL PAPERS INTRODUCTION Defining chronic pain The extent of the chronic pain problem The Biopsychosocial model of pain The neurobiology of chronic pain Neuromatrix Stress and somatic symptoms ICF The diathesis-stress model The chronic pain conditions included in this thesis Psychological factors related to chronic pain Depression Anxiety Catastrophizing Self-efficacy Quality of Life and Health Conclusions of the introduction AIMS OF THE THESIS SUBJECTS AND METHODS Subjects Methods Data collection Statistical analyses RESULTS Study I Study II Study III Study IV DISCUSSION Methodological considerations Data collection Statistics General discussions of the results Generalization of chronic pain Neurobiological mechanisms Prevalence of chronic widespread pain Consequences of widespread pain Widespread pain in a multivariate context Pain intensity and duration The perception of pain intensity Pain intensity related to disability, quality of life and health Pain duration related to disability, quality of life and health Chronic pain and disability
7 8 9 10 10 10 10 11 12 12 14 15 16 17 17 17 18 18 18 19 20 22 22 22 22 27 34 34 37 42 45 49 49 49 50 51 51 51 51 52 52 53 53 53 54 55
Assessment of disability The relationship between disability and depression The relationship between disability and anxiety The relationship between disability and catastrophizing The relationship between disability and self- efficacy Disability and the diathesis-stress model Chronic pain and quality of life and health Assessment of quality of life and health The relationship between HRQL and depression The relationship between HRQL and anxiety The relationship between HRQL and catastrophizing The relationship between HRQL and self-efficacy Chronic pain, HRQL and the ICF model Differences in diagnoses SUMMARY AND CONCLUSIONS FURTHER RESEARCH SAMMANFATTNING PÅ SVENSKA ACKNOWLEDGEMENTS REFERENCES PAPER I-IV
55 55 55 56 56 56 57 57 57 58 58 59 59 59 61 62 63 64 66
ABSTRACT Chronic pain is a very common condition with a prevalence of 40-65% in the community. The high prevalence of chronic pain causes a lot of human suffering but also high societal costs. The development and maintenance of chronic pain constitutes a complex interplay between neurobiological, psychosocial and genetic factors. A biopsychosocial model of chronic pain has been suggested to make a comprehensive context of the understanding of this issue. The main aims of this thesis were to analyze the relationships of the different components of the biopsychosocial model of pain and to study the relative importance of pain, stress and different psychological factors on disability and health related quality of life. The thesis is based on two groups of patients. One group consists of 275 patients with chronic Whiplash Associated Disorder (WAD) and one group comprise 433 patients with WAD, fibromyalgia (FM) and patients with chronic pain related to Spinal Cord Injury (SCI). The patients were investigated by questionnaires assessing different aspects of pain, depression, anxiety, catastrophizing, selfefficacy, disability and Health Related Quality of Life (HRQL). The main results were that psychological factors (especially depression) correlated relatively strongly with perceived HRQL and disability. The degree of depression appeared to have the most important relationship to perceived HRQL. Despite the fact that the patients rated depression just mild or moderate, depression had a great importance for the outcome of HQRL and disability. Pain intensity and duration played, in the cross-sectional perspective, a minor role for perceived HRQL, whereas pain intensity related more to the outcome of perceived disability. From a clinical point of view it is important to assess the complex and unique situation of each individual with respect to depression, anxiety, self-efficacy and pain when planning treatment and rehabilitation in order to optimise the outcome of such programmes.
Arthritis self-efficacy scale
Anxiety Sensitivity Index
Beck Depression Inventory
Central Nervous System
Coping Strategy Questionnaire
Hospital Anxiety and Depression Scale
Health Related Quality of Life
International Classification of Functioning, Disability and Health
Life Satisfaction Questionnaire
Magnetic Resonance Imaging
Non Pain Symptoms Index
Pain Anxiety Symptoms Scale-20
Principal Component Analysis
Pain Catastrophizing Scale
Pain Disability Index
Partial Least Squares by means of Projection to Latent Structures
Pain Regions Index
Quality of Life Scale
Spinal Cord Injury
Visual Analogue Scale
Variable Influence on Projection
Whiplash Associated Disorders
ORIGINAL PAPERS The thesis is based on the following studies, which will be referred to by their roman numerals:
Peolsson M, Börsbo B, Gerdle B. Generalized pain is associated with more negative consequences than local or regional pain: a study of chronic whiplash-associated disorders. J Rehabil Med. 2007 Apr;39(3):260-8.
Börsbo B, Peolsson M, Gerdle B. Catastrophizing, depression, and pain – correlations and influences on quality of life and health: A Study of Chronic Whiplash Associated Disorders (WAD). J Rehabil Med. 2008; 40: 562-569.
Börsbo B, Peolsson M, Gerdle B. The complex interplay between pain intensity, depression, anxiety and catastrophizing with respect to quality of life and disability (Submitted to Disability and Rehabilitation).
Börsbo B, Gerdle B, Peolsson M. Impact of the interaction between selfefficacy, symptoms and catastrophizing on disability, quality of life and health in chronic pain patients (Manuscript).
INTRODUCTION Defining chronic pain The established definition of pain according to IASP (International Association for the Study of Pain) is: ‗An unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage‟. It is noteworthy that pain is always subjective which makes is unquestionable but also limits our ability to assess it with objective methods. It is a sensation in a part or parts of the body, but it is also always unpleasant and therefore as well an emotional experience. Chronic pain is in turn defined as pains persisting over a certain period of time; often 3 or alternatively 6 month are pragmatically used. The patients studied in this thesis, all qualified by far to be classified as patients with chronic pain. The extent of the chronic pain problem. Many studies of the prevalence of chronic pain in the population show that chronic pain is a very common condition. Studies from Sweden found a prevalence of 40 – 65% [1-3] and studies from Great Britain a prevalence of 47 – 51% [4,5]. This high prevalence of chronic pain causes a lot of human suffering but also large society cost. In a Swedish report, the total society cost in Sweden for more serious chronic pain (50-100 mm on the VAS scale) were calculated to 85 billons SEK the year 2003 of which 7% were direct costs for medical care . The Biopsychosocial model of pain The biopsychosocial approach is now widely accepted as a heuristic perspective to the understanding of chronic pain disorders. The biopsychosocial model of pain, illustrated in Figure 2, views physical illnesses such as pain as the result of the dynamic interaction among physiologic, psychological, and social factors, which perpetuates and may even worsen the clinical presentations. Each individual experiences pain uniquely, and a range of psychological and socioeconomic factors can interact with physical pathology to modulate a patient‘s report of symptoms and subsequent disability . In this thesis the biopsychosocial model of pain is used as a structure when planning the investigations and interpreting the results of the studies. The biopsychosocial model of pain can be regarded as a structure that can comprise and include different models and theories. E.g. gate control theory , neuromatrix, the diathesis stress model, ICF and fear avoidance . The biopsychosocial model also includes environmental factors, such as social support, which are of great importance for the individual patients concerning perceived pain and the influence of its consequences . The environmental perspective is however not further investigated in this thesis. The terms theory and model are used in this thesis as they were originally labelled in the literature.
Influence on daily life
Neuromatrix Stress •Acute •Chronic
Nerve impulses from injured, affected or inflammed tissue or nerve
Endokrinologic Immunologic Autonomic
Quality of Life
Figure 2: Diagram illustrating the biopsychosocial model of pain.
The neurobiology of chronic pain The neurobiological systems of nociception and pain are plastic; i.e. when submitted to significant nociception, the function may change in different ways . Chronic pain is a process where both the peripheral and the central nervous system develop an increased sensitivity for different sensory signals (sensitization). Earlier non painful stimuli become painful (allodynia) and/or painful signals are perceived as more painful (hyperalgesia). In acute pain, sensitization is a normal process protecting against more damage. Under certain circumstances this protective mechanism may be over activated and prolonged, i.e. a pathological and noxious sensitization . Several neurobiological processes are involved in the sensitization. In the peripheral nervous system, the nociceptors become more sensitive by pain mediating and pain modulating substances which leads to peripheral sensitization . In the central nervous system several different processes interact to create the central sensitization. Repetitive stimulation of Aδ-fibers leads to a gradual increase of the nerve cell activity in secondary neurons in the dorsal horn of the spinal cord, called „Wind up‟ . In heterosynaptic central sensitization, silent synapses are opened leading to pain produced by low-threshold afferent inputs and the spread of hypersensitivity to regions beyond injured tissue . Long Term Potentiation is coincident activity of pre- and post-synaptic elements, bringing about a facilitation of excitatory input to the dorsal horn and is triggered by short high frequented nociceptiv input [16,17]. There are both inhibitory and facilitating descending pathways for controlling pain transmission . An altered balance between those two can lead to a net facilitation of pain transmission. For an
overview of pain transmission and modulation se Figure 1. This diagram concentrates on the modulation of pain. The pain transmission is far more complicated and includes several other systems and pathways. Neuromatrix The neuromatrix theory of pain  proposes that pain is a multidimensional experience produced by characteristic ‗neurosignature‘ patterns of nerve impulses generated by a widely distributed neural network—the ‗body-self neuromatrix‘— in the brain. These neurosignature patterns may be triggered by sensory inputs, but they may also be generated independently of them. Pain is produced by the output of a widely distributed neural network in the brain rather than directly by sensory input evoked by injury, inflammation, or other pathology. The neuromatrix, which is genetically determined and modified by sensory experience, is the primary mechanism that generates the neural pattern that produces pain. Its output pattern is determined by multiple influences, of which the somatic sensory input is only a part, that converge on the neuromatrix . Stress and somatic symptoms Acute pain activates the HPA (Hypothalamus, Pituary, Adrenal) axis and the sympathetic nerve system. This activation affects the renal and intestinal function as well as the cardiovascular and immune system . Chronic pain is a stressor that will tax the stress system and this prolonged activation of the stress regulation system will ultimately generate breakdowns of muscle, bone, and neural tissue that, in turn, will cause major pain and produce a vicious cycle of pain–stress– reactivity .
Figure 1. Diagram showing pain transmission and modulation pathways in the central nervous system. (Reprinted from Trends in Neurosciences, Vol. 31(4), Zhou M, Cortical excitation and chronic pain, Page 201, 2008  with permission from Elsevier.)
ICF It is important to set chronic pain and persons suffering from it into a larger context than the individual perspective. For this purpose the International classification of functioning, disability and health (ICF) constitutes a useful tool . ICF offers an integrated biopsychosocial model of human functioning and disability and provides a classification system handling several aspects of health and disability. The structure of ICF contains several levels and parts. Figure 3 shows the different levels and parts of the structure. Psychological and physiological functions are included in the component Bodily Function and Structures.
Figure 3. Structure of ICF.
The components in ICF interact with each other (Figure 4). If one component is affected it may modify another component or the health disorder. If body functions and structures are affected, this is referred to as impairment. The reduction of activities is called activity limitation and in participation, a participation restriction. Functioning serves as a sum up term including, body functions, activities and participation. The negative aspect of functioning is disability and includes impairments, activity limitations and participation restrictions. If placing a chronic pain disorder as a health condition in the diagram, it illustrates clearly the great impact on and the great impact of the different components expressing health and disability, including environmental factors. Further, the concept of ICF fits well with the Biopsychosocial model of pain described earlier.
Figure 4. Interactions between the components of ICF.
The diathesis-stress model There are several models and theories, including psychological factors, describing a theoretical framework of the development and maintenance of chronic pain and chronic pain conditions. In this thesis there are several references to the diathesisstress model. This model was first proposed by Gatchel  and developed by Banks and Kerns  to describe the relationship between chronic pain and depression. The diatheses are conceptualized as pre-existing, semi dormant characteristics of the individual before the onset of chronic pain that are then activated by the stress of this chronic condition. The stress component of the model refers to the nature of the chronic pain experience. Banks and Kerns suggest that chronic pain is more likely to result in depression than other chronic medical conditions because of the uniquely challenging nature of stressors associated with chronic pain. These stressors include the aversive sensory and distressful emotional aspects of the pain symptoms, impairment and disability, secondary losses that occur across various domains, and perceived invalidating responses from the medical system. Pincus and Williams  describe the model in a way in which the different components are connected by circular loops and some of the connections are of more importance or have stronger effects than others. Thus, diathesis combines with a primary stressor to produce concurrently a state of emotional distress and disability. The effect of diathesis is to increase perception of pain and enhance distress. Pain and distress affect illness behaviour and increase disability. At this stage disability becomes a stressor in its own right (Figure 5).
Figure 5. The diathesis-stress model with proposed diatheses and negative affect. (Modified from Pincus and Williams ).
The diathesis-stress model is often combined with a epidemiological risk-buffer model or a proactive agentic model. Hence, protective factors, such as selfefficacy can buffer the adverse effects of stressors or act as a factor for proactive shaping of life circumstances . The chronic pain conditions included in this thesis Whiplash Associated Disorders (WAD) A whiplash injury is an injury caused by a whiplash trauma. This trauma implies a strain on the structures of head and neck when a acceleration-deceleration movement transfers forces to those structures without a direct trauma to the head and neck. The majority of subjects with acute WAD will be cured within three months after the trauma , but for a minority, the acute neck pain may develop into a chronic pain condition. The prevalence of chronic pain in the community as a consequence of whiplash trauma has been estimated to 1-1.5% [30,31]. Fibromyalgia Fibromyalgia (FM) is a subgroup of chronic generalized pain disorders. It is diagnosed according to the criteria proposed by American Collage of Rheumatology (ACR) . Long term multifocal muscle pain and widespread allodynia, mainly pressure allodynia, are compulsory symptoms. The prevalence in the population is suggested to 2-4% .
Chronic pain associated to Spinal Cord Injury Chronic pain is an important problem following Spinal Cord Injury (SCI). The reported prevalence varies but averages 65% with around one third rating their pain as severe . A number of specific types of SCI pain can be distinguished based on descriptors, location and response to treatment . Nociceptiv pain can arise from musculoskeletal structures and viscera and neuropathic pain can arise from spinal cord and nerve damage. About 50% of the pain is considered to be neuropathic pain. Psychological factors related to chronic pain Depression The prevalence of depression in the general population is estimated to 4-10% . Depression is not simply a comorbid condition but interacts with chronic pain to increase morbidity and mortality. High frequencies of depressive symptoms e.g. 34.8% in FM, have been found in patients with chronic pain [26,36]. Depressed chronic pain patients report greater pain intensity, greater interference from pain, more pain behaviours, less life control, and more use of passive/avoidance coping strategies than chronic pain patients without depression [37,38]. The temporal relationship between chronic pain and depression is under debate. Fishbain et al.  found strong support for the consequence hypothesis: depression is a consequence that follows the development of pain. To describe the relationship between chronic pain and depression, Banks and Kerns  suggested a diathesis-stress-model where the diathesis is described as pre-existing, semi-dormant characteristics of the individual before the onset of chronic pain. These characteristics are activated by the stress of the chronic condition and may lead to depression. Qualitative differences between depression as a result of chronic pain and depression as a primary psychiatric disorder have been reported [39,40]. Pincus and Morley suggest that ―affective distress‖, which incorporates wider emotions such as anger, frustration, fear, and sadness, is a better term than depression . Depression is a predictor of disability in chronic pain patients in long-time follow-up studies . There is also a relationship between depression and poorer self-reported functional activity among persons with chronic pain . Anxiety The prevalence of anxiety disorders in the general population is estimated to 1217% . Anxiety is a co-morbidity to chronic pain with incidence rates between about 15 and 40% , co-morbidity also exist between mood and anxiety disorders [45,46]. Several studies have found pain conditions being more strongly associated with several anxiety disorders than with depression [47,48]. Patients with anxiety disorders reported the highest pain intensity and interference and the lowest general activity level in fibromyalgia . Pain related anxiety includes physiological, cognitive, behavioural and affective manifestations of anxiety within the context of pain . Heightened levels of anxiety about pain are believed to contribute to avoidance of activities that are perceived to promote pain, which in turn, often lead to physical deconditioning, secondary behavioural problems and reduced social contact . This pattern of responding is likely to become cyclic in nature, such that emotional responsivity and physical deconditioning lead to greater levels of pain, behavioural 17
interference, perceived lack of control over life activities and affective distress [51,52]. In this model, anxiety about pain is a critical psychological factor involved with the production of maladaptive responding, behavioural interference, and emotional distress. Anxiety sensitivity is the fear of arousal-related bodily sensations arising from beliefs that these sensations have harmful consequences, a catastrophically misinterpretation [52,53]. Anxiety sensitivity has been closely associated with negative pain experiences in acute and chronic settings [54,55]. Catastrophizing Catastrophizing has broadly been defined as an exaggerated negative orientation toward pain stimuli and pain experience . Studies have indentified relationships between catastrophizing and psychological distress , physical functioning and disability , ratings of pain intensity , interference with life activities , psychosocial dysfunction  and quality of life . Knowledge about whether catastrophizing is a cause or a consequence of chronic pain is still lacking . Studies that can be interpreted in either of these directions do however exist [56,64-67]. Self-efficacy Perceived self-efficacy is defined as people's beliefs about their capabilities to produce designated levels of performance that exercise influence over events that affect their lives . Self-efficacy beliefs regulate human functioning through cognitive, motivational, affective, and decisional processes . These beliefs affect whether individuals think in self-enhancing or self-debilitating ways, how well they motivate themselves and persevere in the face of difficulties, the quality of their emotional life, and vulnerability to stress and depression. Research verifies the predictive generality of efficacy beliefs as significant contributors to the quality of human functioning . People with a high sense of coping efficacy adopt strategies and courses of action designed to change hazardous environments to benign ones. In this mode of affect regulation, efficacy beliefs alleviate stress and anxiety by enabling individuals to mobilize and sustain coping efforts. Selfefficacy operates as a cognitive regulator of stress and anxiety arousal . Several studies have noted that high scores on self-efficacy are inversely related to pain intensity. This is relevant for different pain conditions such as arthritis , musculoskeletal pain , cancer pain , headache , pain in SCI , and other chronic pain . Self-efficacy is also inversely connected to depression [77,78] and is a good predictor for pain-related disability [79,80].
Quality of Life and Health There are several and still evolving definitions of quality of life, health and life satisfaction. It is beyond the scope of this thesis to evaluate and elaborate on those definitions. In the following is accounted for the concepts used in this thesis. Revicki et al  define QOL as ‗a broad range of human experiences related to one's overall well-being. It implies value based on subjective functioning in comparison with personal expectations and is defined by subjective experiences, 18
states and perceptions. Quality of life, by its very natures, is idiosyncratic to the individual, but intuitively meaningful and understandable to most people‘ (p. 888). This definition denotes a meaning for QOL that transcends health, and it can be argued that ―health‖ can be included in this definition. Health related Quality of Life (HRQL) is denoted as the subjective assessment of the impact of disease and treatment across the physical, psychological, social and somatic domains of functioning and well-being . Life satisfaction is often measured within the same domains as QOL, but with the distinction that this concept focuses on the individual's perception of the difference between the subjective reality and needs or wants. Those types of instruments can therefore be considered to be associated with of the HRQL family of instruments . In this thesis all instruments measuring QOL (SF-36, EuroQol, LiSat-11, and QOLS) are used on patients with a chronic health condition, and are therefore considered to evaluate Health Related Quality of Life and in a broader sense the aspect of health. The concept general health is used when evaluating the subscale ―general health‖ in SF-36.
Conclusions of the introduction There are a great number of studies addressing the importance of isolated psychological factors in chronic pain [77,84] but a much lesser number of studies investigating the interrelationship with psychological factors and pain variables and there relative importance on different outcome variables. Based on the literature it is reasonable to expect that patients with e.g. high pain intensity , depression  and catastrophizing  will perceive health and quality of life considerably worse than those patients rating their situations with respect to these three factors better. Using e.g. certain regression techniques the mean influences of these three factors upon health and quality of life can theoretically be determined at group level for each outcome variable separately. However, the clinical question might be more complex; e.g., is the effect of high catastrophizing with respect to health and quality of life similar when e.g. pain intensity is high or low? Or is it from a treatment or rehabilitation perspective important to intervene against high catastrophizing regardless of pain intensity in patients with chronic pain?
AIMS OF THE THESIS The overall aims of this thesis were:
To analyze how different components of the bio- psycho- social model of pain interacts.
To study the relative importance of pain, stress and different psychological factors on daily life, disability, health, life satisfaction and quality of life.
To, from the above mentioned aspects, compare chronic pain conditions of different origin and character.
Specific aims of the different studies where:
Study I This study was made to answer the following questions:
Does chronic WAD with widespread pain have more severe consequences with respect to other symptoms, coping strategies, and different aspects of perceived health than chronic WAD patients with local/regional pain? Do pain, depression, and not directly pain related symptoms intercorrelate and to what extent do these symptoms correlate with catastrophizing?
To classify subgroups according to the degree of pain intensity, depression, and catastrophizing and to investigate the distribution in a group of chronic WAD patients. To investigate how these subgroups were distributed and interrelated multivariately with respect to consequences such as health and quality of life outcome measures.
To identify subgroups based on the occurrence of depression, anxiety, catastrophizing and the degree of pain intensity and duration and To investigate how the subgroups differed with respect to background variables, diagnosis, pain related disability and perceived quality of life.
Study IV To investigate the interaction between self-efficacy, including subcomponents, and symptoms (pain, depression, and anxiety), catastrophizing, disability, quality of life and health in a population of chronic pain patients.
SUBJECTS AND METHODS Subjects This thesis is based on two different groups of patients with chronic pain. Group 1 (Study I and II) All patients came from the consecutive flow of patients seeking care at the Pain and Rehabilitation Centre of the University Hospital, Linköping, Sweden and this cross-sectional study is based on 275 patients. Patients fulfilling the criteria of WAD grades II or III according to the Quebec classification , were included in the study. Clinical examination and case history established diagnoses of chronic WAD. Radiological evaluation (X-ray, MRI) was only performed when there was a suspicion of skeletal damage or disc herniation. Group 2 (Study III and IV) 891 patients at the clinical rehabilitation departments at Linköping University Hospital and County Hospital Ryhov in Jönköping from 2002 through 2004 were invited to participate. The inclusion criteria were chronic pain (>3 month), and one of the following diagnoses: Spinal Cord Injury (SCI) related pain, Fibromyalgia (FM), and Chronic Whiplash Associated Disorders (WAD). The patients were selected from the medical records. The diagnoses, settled by experienced clinicians, where obtained from the medical records. Patients with double diagnoses were excluded. Out of 891 invited patients, we received 434 returned questionnaires after two reminders. One patient did not satisfy the inclusion criteria and was excluded. Thus a total of 433 patients – covering 47 patients with SCI-related pain, 150 with WAD, and 236 with FM – constituted the data material in the second study group. Methods Data collection Group 1 Each patient received a questionnaire shortly before the examination at the centre. The questionnaire was completed at home and was delivered to the physician at the visit to the centre. The questionnaire contained the following items and instruments: Age, gender, and background data. Number of days on sick-leave during the previous12 months, number of months out of occupation, degree of sick leave (0%, 25%, 50%, 75%, or 100%), degree of disability pension (0%, 25%, 50%, 75% or 100%), and number of visits to physician in the recent 6 months. Pain intensity ratings at 11 predefined anatomical regions. For the rating of pain intensity, a visual analogue scale (VAS) was used; the scale was a 100 mm long with defined end points (‘no pain’ and ‘worst pain imaginable’) but
without marks in between (results in cm). All the questions regarding pain concerned the previous 7 days [87,88]. Pain Regions Index (PRI). Number of the above pre-defined anatomical regions associated with pain with a possible range of 0-11.
Figure 6. Diagram shown in the questionnaire, to define the anatomical regions
Presence of other pain related symptoms: radiating pain to the arm(s), radiating pain to the leg(s), headache, perception of heavy head and pain in the throat. For each of these symptoms, the patients chose among the following alternatives: 0 = ‘no, never’; 1 = ‘no, seldom’; 2 = ‘yes, occasionally’; 3 = ‘yes, often’. In the analyses and tables these symptoms were dichotomised (0-2 versus 3). The Beck Depression Inventory (BDI) evaluates 21 symptoms of depression into a scale ranging between 0 to 63. Scores of less than 10 indicate no or minimal depression, 10-18 indicate mild to moderate depression, 19-29 indicate moderate to severe depression, and scores of 30 or more indicate severe depression. For psychiatric patients, a screening cut-point of 12/13 is suitable, whereas 9/10 is appropriate in screening medical patients (used in the present study) . Thirty-one different symptoms - not directly pain related - were registered: sleeping difficulties, tachycardia, bowel problems, gastritis, fatigue-tiredness, weak voice, nausea, anxiousness, difficulties with changes in light intensity, concentration problems, hoarseness, difficulties with swallowing, difficulties with urinating, vertigo, numbness in hands, changed perception hands, blurred vision, defecation problems, sound sensitivity, changes in alcohol sensitivity, light sensitivity, feeling of fullness of ear, irritable, memory problem, diminished field of sight, low mood, changed perception of touch legs, difficulties with control of legs, fatigue in legs, twitches in the legs, and difficulties walking down in stairs. For each symptom the patients chose among the following alternatives: 0 = ‘no, never’; 1 = ‘no, seldom’; 2 = ‘yes,
occasionally’; 3 = ‘yes, often’. In the analyses and tables, these symptoms were dichotomised (0-2 versus 3). Non Pain Symptoms index (NPSI) An index that counted the number of the above not directly pain related 31 symptoms (in the dichotomized form) was also computed with a possible range: 0-31). Coping Strategy Questionnaire (CSQ) is frequently used to measure how patients cope with pain and includes eight types of coping strategies with the aim to describe how patients cope with pain. These coping strategies are diverting attention, re-interpreting pain sensation, coping self-statements, ignoring pain sensations, praying and hoping, catastrophizing, increased behavioural activities, and pain behaviour. Each strategy is evaluated according to its frequency of use, ranging from never (0) to always (6) with a maximum score of 36. Two additional questions concern the perception of control and possibility to minimize pain (not used in the present study). The Swedish version of the Coping Strategy Questionnaire (CSQ)was used in the present study . Life Satisfaction Questionnaire (LiSat-11) consisted of estimations of life satisfaction in general as well as 10 specific domains to be estimated: satisfaction with vocational situation, financial situation, leisure situation, contact with friends and acquaintance, sexual life, ADL, family life, and partnership. Two additional variables were added to this list – satisfaction with physical and psychological health. Each item has six possible answers: 1 = very dissatisfying; 2 = dissatisfying; 3 = fairly dissatisfying; 4 = fairly satisfying; 5 = satisfying; 6 = very satisfying . SF-36 Health Survey (Swedish version) is an instrument that intends to represent multi-dimensional health concepts and measurements of the full range of health states, including levels of well-being and personal evaluations of health. The instrument covers 36 questions covering 8 dimensions: physical functioning (SF 36pf), role limitations due to physical functioning (SF 36rp), bodily pain (SF 36bp), general health (SF 36gh), vitality (SF 36vit), social functioning (SF 36sf), role limitations due to emotional problems (SF 36re), and mental health (SF 36mh). Each item score is coded, summed, and transformed to a standardized scale calculated from a specific score algorithm (ranging from 0 – 100 with two end points identified as ―worst‖ and ―best‖ possible health state. The transformed score has been used in this study . EuroQol instrument captures a patient‘s perceived state of health. A state of health is defined as combinations of five dimensions and three levels of choice (no problems, some problems, or severe problems) for each dimension: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. This descriptive system covers the first part of the instrument. The answers are coded (1-3). The codings are transformed by a table or by using an algorithm to score the findings (EQ-5D). A second part concerns a self-estimation of today‘s health according to a 100-point scale, a ‗thermometer‘ (EQ-VAS) with defined end points (high value indicates good health and low value indicates bad health). Thus the two parts comprise different aspects related to health as quality of life. In this study the total score (EQ-5D) and the self-estimation scale (EQ-VAS) are reported .
Group 2 The patients were asked by a letter to participate and the patients who chose to participate received a postal questionnaire covering background data, psychological and health-related items. Patients who did not return the questionnaire were reminded twice before they were indicated as dropouts. The questionnaire included the following items and instruments. Swedish validated versions were used; references given below present the questionnaires and studies of psychometrical properties: Age, gender and background data Pain intensity ratings of 9 predefined anatomical regions. For the rating of pain intensity, a visual analogue scale (VAS) was used; the scale was a 100 mm long with defined end points (―no pain‖ and ―worst pain imaginable‖), but without marks in between (results in cm). All the questions regarding pain concerned the previous 7 days. The rating of the most painful region was used (VAS-max) [87,88]. Pain Regions Index (PRI). Number of the above pre-defined anatomical regions associated with pain with a possible range of 0-9. Anxiety Sensitivity Index (ASI) is a 16-item self-reported questionnaire. Each item asks about the amount of fear the participant experiences in regard to bodily sensations commonly associated with anxiety. Participants are asked to rate each item on a 5-point Likert-like scale ranging from very little (0) to very much (4). The ratings on the 16 items are summed for a total ranging from 0 to 64. Studies have found support for test-retest reliability, criterion validity and construct validity (e.g., support for the distinction between AS and trait anxiety) [53,94]. Pain Anxiety Symptoms Scale-20 (PASS-20) is a short version of the 40-item PASS that measures fear and anxiety responses specific to pain. The PASS-20 has four 5-item subscales that measure Avoidance, Fearful thinking, Cognitive anxiety and Physiological Responses to Pain. Participants rate each item on a 6-point scale ranging from never (0) to always (5). Reliability analyses with PASS-20 indicate good internal consistency akin to the PASS-40. Psychometric analyses reveal good convergent, discriminant, predictive and construct validity [95,96]. Hospital Anxiety and Depression Scale (HADS) is a self-rating scale in which the severity of anxiety and depression is rated on a 4-point scale. Seven questions are related to anxiety and seven to depression, both with a score range of 0–21. A score of 7 or less indicates a non-case, a score of 8–10 a doubtful case, and 11 or more a definite case. The instrument is widely used in clinical practice and research. Investigations have shown that the HADS is a psychometrically sound instrument. In this study, we used both subscales [97,98]. The Pain Catastrophizing Scale (PCS) is a 13-item self-report measure designed to assess catastrophic thoughts or feelings accompanying the experience of pain. Respondents are asked to reflect on past painful experiences and to indicate the degree to which each of the 13 thoughts or feelings are experienced when in pain. The questionnaire uses a 5-point scale ranging from 0 (not at all) to 4 (all the time). Subscales for rumination, magnification and helplessness plus a total score are added up. In this study, we used the total score [56,99]. 25
Quality of Life Scale (QOLS-S) is composed of 16 items that together describe the quality of life concept: (i) Material comforts; (ii) Health; (iii) Relationships with parents, sibling and other relatives; (iv) Having and rearing children; (v) Close relationships with spouse or significant others; (vi) Close friends; (vii) Helping and encouraging others, participating in organizations, volunteering; (viii) Participating in political organizations or public affairs; (ix) Learning; (x) Understanding yourself; (xi) Work; (xii) Expressing yourself creatively; (xiii) Socializing; (xiv) Reading, music or watching entertainment; (xv) Participating in active recreation; and (xvi) Independence, being able to do things for yourself. A seven-point satisfaction scale is used. Clients estimated their satisfaction with their current situation. A higher total score shows higher satisfaction. The item scores are added to a total score, ranging from 16 to 112 [100-103]. SF-36 Health Survey (Swedish version) see group 1. The Pain Disability Index (PDI) is a 7-item self-report instrument based on a 10-point scale that assesses perception of the specific impact of pain on disability that may preclude normal or desired performance of a wide range of functions, such as family and social activities, sex, work, life-support (sleeping, breathing, eating), and daily living activities. The PDI has shown good reliability and validity in several studies [72,104]. The arthritis self-efficacy scale (ASES) is a standardized questionnaire with 20 items which measure an individual‘s perceived self-efficacy to cope with the consequences of chronic arthritis. In this study a validated Swedish version for chronic pain was used. The only modification made was to change the words ´arthritis pain‘ and ‗arthritis‘ to ‗pain‘. The first five-item subscale assesses self-efficacy perception for controlling pain (SE-pain). The second nine-item subscale assesses self-efficacy for performing functions in daily living (SE-function). The six-item subscale measures self-efficacy for controlling other symptoms related to chronic pain (SE-symptom). Each question is followed by a scale for, marking the answer, from 10 to 100. Each subscale is scored separately, by taking the mean of the subscale items [105,106].
Table 1: Summary of variables and instruments used in the different studies.
x x x
x x x
x x x
x x x x x
x x x x
x x x
x x x
Items/Instrument VAS Pain duration PRI Symptoms NPSI Depression BDI HADS-D Anxiety HADS-A PASS-20 ASI Catastrophizing CSQ PCS Self-efficacy ASES Quality of Life LiSat-11 EuroQol QOLS-S Health SF-36 Disability PDI
Statistical analyses All statistical evaluations were made using the statistical packages SPSS (versions 12.0 and 15.0) and SIMCA-P+ (versions 10.2 and 11.1). Results in text and tables are generally given as mean values ± one standard deviation (±1SD). In all statistic analyses, p ≤ 0.05 was regarded as significant. The statistical analyses used in the different studies are given in Table 2.
Table 2. Statistical analyses used in studies I-IV. ANOVA = Analysis of variance; PCA = Principal Component Analysis. PLS = Partial Least Squares by means of Projection to Latent Structures.
Methods/Study Krauskal-Wallis Mann Whitney Spearman´s rho Chi2 ANOVA Cluster analysis PCA PLS
Study I x x x x
x x x x x
Multivariate projection analyses were made using the software SIMCA P+ 11,5. In all studies Principal Component Analysis (PCA) and Partial Least Squares by means of Projection to Latent Structures (PLS)were used . The basic analytic questions to which these projection methods can be applied are (a) overview of data, (b) classification and/or discrimination among groups of observations and (c) regression modelling between two blocks of data (X) and (Y). The multivariate methods (using PCA and PLS) are inductive statistical methods where all variables are included. The variables are mean-centred and given the same variance and thus given the same possibility to influence the model. The variables create a model where the interrelationships between the variables, the subjects and between subjects and variables become highlighted. The basic aim applying these projection methods is that data can be reduced to a few latent variables that summarize the original variables. Other advantages are that they do not require interval scaled data, copes with multicollinearity, copes with missing data, is robust to noise in both X and Y and can be used with small samples, even with more original variables than subjects. Principal Component Analysis (PCA) In the PCA the overall pattern of correlations between variables and observations (subjects) may be visualized and at the same time summarised by latent variables. Each variable defines a co-ordinate axis. If the data material has got K variables, exemplifies a K-dimensional space has to be handled. Since a multidimensional space is hard to imagine, figure x illustrates a three-dimensional space with the variables x1, x2 and x3. Each subject is defined by a point in the coordinate system. (Figure 7)
Figure 7: A co-ordinate system with three axes representing variables x1, x2 and x3. Each subject is defined by a point in the three-dimensional space. Modified with permission from Eriksson et al .
After mean centering (Figure 8), the first principal component is calculated by way of the least square method. It is placed in the direction of the largest variation (Figure 9).
The second principal component is orthogonally projected to the first principal component along the second largest variation. The two principal components create a plane on witch the observations are projected (Figure 10). A component consists of a vector of numerical values between -1 and 1, referred to as loadings and obtained significant components are uncorrelated. Variables that have high loadings (with a positive or negative sign) on the same component are inter-correlated. Items with high loadings (ignoring the sign) are considered to be of large or moderate importance for the component under consideration. Items with high absolute loadings on a component but with different signs are negatively correlated.
Figure 8: Mean centering.
Figure 9: The first principal component calculated.
Figure 10: PC 1 and PC 2 in relation to the original variables.
Two plots are generated from the PCA analysis: the loading plot (Figure 11) describes the correlations between the variables; while the score plot (Figure 12) describes the correlations between the subjects. Subjects clustered together share similar characteristics, whereas the characteristics of subjects far from each other differ. Since the two plots are complementary, e.g. subjects located far to the right horizontally in the score plot have high values on the variables far to the right in the loading plot.
Figure 11: The loading plot shows the relationships between the variables.
Figure 12: The score plot shows the relationships between the subjects.
Projections to Latent Structures by means of Partial Least Squares (PLS) PLS is used to regress one or several Y-variables using several other variables (Xvariables)  and calculates the covariance between a set of X-variables and a set of Y-variables. In PLS the principal components are projected based on the same techniques as in PCA. PLS is useful to predict an outcome. PLS provides variable-related parameters to facilitate the interpretation of the model. Regression coefficients are used to obtain detailed information whether the variable had a significant positive or negative impact as well as the magnitude on the regressed outcome variable (Figure 13). The statistical significance of each coefficient is indicated as 95% confidence interval not including zero. The benefit of this procedure is to provide a single vector of concise model information per response variable. The disadvantage is that the correlation structure among the responses is lost, but this relationship has already been elaborated according to the PCA plot and the PLS table.
lCoeffCS[Last comp.] Y-variabel 1
Var ID (Primary) SIMCA-P+ 11.5 - 2007-02-21 10:52:59
Figure 13: Graphic illustration of regression coefficients in PLS.
The VIP variable (variable influence on projection) gives information about the relevance of each X-variable and each Y-variable pooled over all dimensions (Figure 14). VIP is a weighted sum of squares of the PLS weights. Because the weights express the correlation between the X and Y matrices, they summarize the importance of the X-variables. The PLS regression coefficients may be reexpressed as a regression model and express the influence of each X-variable on Y in each single component. The variable of importance for explaining Y is primarily identified by a VIP value ≥1.0 and secondary by the regression coefficient in relation to Y. VIP[Last comp.] 1,6 1,5 1,4 1,3 1,2 1,1 1,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2
Var ID (Primary) SIMCA-P+ 11.5 - 2007-02-21 10:57:04
Figure 14: Graphic illustration of the VIP variable in PLS.
Two concepts – R2 and Q2 – are further used to describe the results of the modelling procedure in PCA and PLS. R2 describes the goodness of fit: the fraction of sum of squares of all the variables explained by a principal component as is given both for X-variables and Y-variables, i.e. the degree of the variation in X which is used to explain the variation in Y. Q2 describes the goodness of prediction: the fraction of the total variation of the variables that can be predicted by a principal component using cross validation methods. Outliers are identified using the two powerful methods available in SIMCA-P: score plots in combination with Hotelling‘s T2 (identifies strong outliers) and distance to model in X-space (DModX) (identifies moderate outliers).
RESULTS Study I Formation of three subgroups based on number of regions with pain (PRI) The group of patients with chronic WAD was divided into three subgroups based on PRI (group 1: 0-3 regions (n=45; 17%); group 2: 4-7 regions (n=152; 56%); and group 3: 8-11 regions (n=74; 27%). Very similar results with respect to the different consequences reported below were obtained when cluster analysis was made. Background data Somewhat (non-significant) higher proportions of women were found in the group with highest PRI (group 3). There was a significant difference in degree of disability pension between groups (highest prevalence in group 3). Pain symptoms Pain intensity in the head, the neck and shoulders, and the low back differed significantly between the three groups. The frequencies of radiation to arms and legs were highest in group 3 and lowest in group 1. No significant differences were found for the other pain related symptoms. Not directly pain related symptoms including depression In the whole cohort, the proportion with >10 on BDI (i.e., at least mild to moderate depression) was 55.9% and with >19 on BDI (i.e., at least moderate to severe depression) was 23.6%. BDI differed between the three groups (p10 on BDI were 36.4% in group 1, 68.7% in group 2, and 78.1% in group 3. Corresponding figures for >19 on BDI were 2.3%, 24.7%, and 34.2%. The NPSI differed significantly between the three groups (p=0.027); group 1 had lowest number of symptoms. The individual items that differed between the three groups were sleeping difficulties (p = 0.020), fatigue-tiredness (p = 0.014), and memory problems (p = 0.043). Coping Strategy Questionnaire (CSQ) Significant differences were found for two of the subscales of CSQ. Group 3 with high PRI showed highest values both for the catastrophizing (p = 0.013) and the reinterpret pain sensation (p = 0.023) subscales and group 1 (low PRI) had the lowest values. Aspects of life satisfaction and generic health Eight out of 11 scales of the life satisfaction instrument LiSat-11 showed significant differences between the three groups; no significant differences were found for the subscales vocational situation, sexual life, and partnership relations. According to SF-36, all scales except ―Role physical‖ differed between the three groups even though a similar trend for the other scales was found. EQ-5D (p